One disadvantage of shrinking microelectronics is that the testing of integrated circuits becomes exponentially more complex. A convenient and efficient means of testing increasingly complex integrated circuits is to design them with testing hardware already built-in. In 1985, a consortium of European electronics companies formed the Joint Test Action Group (JTAG) in order to create an industry standard which would define an architecture and software protocol for built-in testing hardware. The resulting standard, sometimes referred to as JTAG, was adopted by the Institute of Electrical and Electronics Engineers (IEEE) as IEEE Standard 1149.1.
The JTAG standard uses a boundary-scan architecture, so called because the JTAG circuitry is connected to a dedicated path around the integrated circuit's boundary (where the I/O pins are located). JTAG uses this boundary path to directly input to, and download data and instructions from, the I/O pins of the integrated circuit. The JTAG capability of an integrated circuit is accessed through a multi-pin JTAG test access port (TAP), through which instructions and data from external testing software are input and instructions and data from the integrated circuit are output.
The JTAG interface and circuitry can be used to implement debugging and emulation functions. In debugging embodiments, the JTAG interface is used to download code, execute it, and examine register and memory values. In system-on-a-chip (SOC) embodiments which perform emulation, an in-circuit emulator (ICE) is often embedded in the SOC. The embedded ICE is an actual processor, or a specially modified version of the standard processor, combined with special purpose instrumentation logic to perform the debugging and emulation functions. With emulation, one can test the embedded processor software with techniques such as code trace, benchmarking, trapping, evaluating breakpoints, profiling, etc. However, such extra functions usually require extra pins (besides the JTAG TAP) to output the resulting data.
FIG. 1 shows a conventional SOC 100 with an embedded ICE 105 located within the embedded processor 110 of the SOC 100. An external debugging and emulation program, located in personal computer (PC) 120, uses the JTAG interface logic 125 (by means of JTAG TAP, not shown here) to access and control ICE 105. When running, ICE 105 takes over the embedded processor 110 which controls most of the functions of the SOC 100. As such, ICE 105 can access, control, upload, and download from any of the memory 130 or peripheral modules 135 which control other SOC components 137.
The output of the debugging and emulation is forwarded to trace logic 140 and, more particularly, to trace FIFO (First-In, First-Out) buffer 145, where the trace output is collected for download, through trace port and status analyzer 150, to the external debugging and emulation program on PC 120, where the trace output is analyzed.
One problem with the prior art design is that the trace port and status analyzer 150 uses multiple pins of the SOC for the parallel output of the trace data. The amount of data is thus limited by the number of pins which can be designated as trace output pins in the SOC design. As SOCs become more complex and faster, not only does the total number of I/O pins needed for critical functions increase—thus reducing the number of available pins which can be designated as trace output pins—but also the amount of trace data increases, thus requiring more trace output pins for transferring the increased trace data. Furthermore, bandwidth limitations may limit data rates between the trace port and status analyzer 150 and the on chip trace logic 140.
Usually, it is the number of trace output pins that are sacrificed for other uses in integrated circuit design, thereby causing difficulties as well as a bottleneck at the trace port output of the integrated circuit. One solution to this problem is to create a larger trace FIFO buffer 145 so that more data may be stored while the parallel output is downloading as much data as it can. However, area on an integrated circuit is in short supply, and using up more area for a trace FIFO buffer is expensive and wasteful.